@article{ART003397369},
author={Jisu Sohn and Kyuhwan Lee and YoungGeun Park and Dongsu Lee},
title={Generative AI-Based ITGC Audit Workpaper Automation: Performance Evaluation and Practitioner Acceptance},
journal={Journal of The Korea Society of Computer and Information},
issn={1598-849X},
year={2026},
volume={31},
number={10},
pages={47-69}
TY - JOUR
AU - Jisu Sohn
AU - Kyuhwan Lee
AU - YoungGeun Park
AU - Dongsu Lee
TI - Generative AI-Based ITGC Audit Workpaper Automation: Performance Evaluation and Practitioner Acceptance
JO - Journal of The Korea Society of Computer and Information
PY - 2026
VL - 31
IS - 10
PB - The Korean Society Of Computer And Information
SP - 47
EP - 69
SN - 1598-849X
AB - Reliability concerns arising from hallucinations and automation bias constrain the adoption of generative AI in auditing. This study proposes an automated model for drafting Information Technology General Controls (ITGC) workpapers using an eight-block prompt and type-safe output. The system was compared with an experienced evaluator using identical evidence across six controls, supplemented by a practitioner survey. In the initial single-run experiment, workpaper drafting time decreased by 96.2%, and agreement with the single reference evaluator reached 98.7% across 154 refined synthetic-evidence cases; because 148 of the 154 cases were PASS, this figure reflects agreement on a highly imbalanced dataset rather than general audit accuracy, and the detection of all four FAIL cases is a preliminary result. No hallucinations were identified under the predefined criterion in the final post-processed outputs. However, agreement decreased to 83.6% in a subsequent rerun of 146 cases across three controls. Three evaluators showed complete agreement on 60 cases (Fleiss’ κ = 1.00). The experienced evaluator performed better in identifying exceptions requiring contextual judgment, while the survey indicated perceived needs for human intervention. These findings are observations under refined synthetic evidence and the specific system configuration examined; they do not establish that individual system components caused the high agreement or the absence of hallucinations, and the effects of human intervention on error reduction and workpaper quality require further validation.
KW - Generative Artificial Intelligence;Internal Control over Financial Reporting (ICFR);IT General Controls (ITGC);Professional Skepticism;Hallucination;Human-in-the-Loop (HITL)
DO -
UR -
ER -
Jisu Sohn, Kyuhwan Lee, YoungGeun Park and Dongsu Lee. (2026). Generative AI-Based ITGC Audit Workpaper Automation: Performance Evaluation and Practitioner Acceptance. Journal of The Korea Society of Computer and Information, 31(10), 47-69.
Jisu Sohn, Kyuhwan Lee, YoungGeun Park and Dongsu Lee. 2026, "Generative AI-Based ITGC Audit Workpaper Automation: Performance Evaluation and Practitioner Acceptance", Journal of The Korea Society of Computer and Information, vol.31, no.10 pp.47-69.
Jisu Sohn, Kyuhwan Lee, YoungGeun Park, Dongsu Lee "Generative AI-Based ITGC Audit Workpaper Automation: Performance Evaluation and Practitioner Acceptance" Journal of The Korea Society of Computer and Information 31.10 pp.47-69 (2026) : 47.
Jisu Sohn, Kyuhwan Lee, YoungGeun Park, Dongsu Lee. Generative AI-Based ITGC Audit Workpaper Automation: Performance Evaluation and Practitioner Acceptance. 2026; 31(10), 47-69.
Jisu Sohn, Kyuhwan Lee, YoungGeun Park and Dongsu Lee. "Generative AI-Based ITGC Audit Workpaper Automation: Performance Evaluation and Practitioner Acceptance" Journal of The Korea Society of Computer and Information 31, no.10 (2026) : 47-69.
Jisu Sohn; Kyuhwan Lee; YoungGeun Park; Dongsu Lee. Generative AI-Based ITGC Audit Workpaper Automation: Performance Evaluation and Practitioner Acceptance. Journal of The Korea Society of Computer and Information, 31(10), 47-69.
Jisu Sohn; Kyuhwan Lee; YoungGeun Park; Dongsu Lee. Generative AI-Based ITGC Audit Workpaper Automation: Performance Evaluation and Practitioner Acceptance. Journal of The Korea Society of Computer and Information. 2026; 31(10) 47-69.
Jisu Sohn, Kyuhwan Lee, YoungGeun Park, Dongsu Lee. Generative AI-Based ITGC Audit Workpaper Automation: Performance Evaluation and Practitioner Acceptance. 2026; 31(10), 47-69.
Jisu Sohn, Kyuhwan Lee, YoungGeun Park and Dongsu Lee. "Generative AI-Based ITGC Audit Workpaper Automation: Performance Evaluation and Practitioner Acceptance" Journal of The Korea Society of Computer and Information 31, no.10 (2026) : 47-69.
